CN102436651B - Method and system for extracting three-dimensional hierarchical boundaries of retina OCT (Optical Coherence Tomography) volumetric data - Google Patents

Abstract

The invention discloses a method and a system for extracting three-dimensional hierarchical boundaries of retina OCT (Optical Coherence Tomography) volumetric data, and relates to the technical field of medical images. The method comprises the following steps: S1, carrying out vertical difference filtering operation on the retina OCT volumetric data to obtain difference volumetric data; S2, calculating to obtain boundary index volumetric data; S3, searching in the boundary index volumetric data to obtain divergent boundary points of a current boundary; S4, smoothing the set of the divergent boundary points to obtain the current boundary; S5, judging whether all boundaries of the retina OCT volumetric data are obtained, if yes, carrying out step S6, or else, updating the retina OCT volumetric data according to the current boundary and updating convolution operators of the difference filtering operation, and returning to the step S1; and S6, outputting all the boundaries of the retina OCT volumetric data. The method and system disclosed by the invention have the advantages of full automation, accuracy, stability, high efficiency and strong applicability.

Description

Extracting method and the system on the three-layer laminated border of retina OCT volume data

Technical field

The present invention relates to the medical image technical field, particularly extracting method and the system on the three-layer laminated border of a kind of retina OCT volume data.

Background technology

Optical coherence tomography (Optical Coherence Tomography, OCT) promoted rapidly in recent years, obtained to use more and more widely at medical domain, its image-forming principle is by the light intensity of measuring object back scattering or reflection, it to be carried out fault imaging.The OCT technology has the significant advantage such as real-time, non-destructive, high resolving power, can generate fast the full resolution pricture of organization internal.OCT is mainly used in the diagnosis of the diseases such as ophthalmology disease and coronary artery at present, and retina OCT image is the result that OCT equipment obtains the retina image-forming in eyeball.

But exist the much noise that causes due to various factors in the OCT image, the speckle noise that especially causes due to the coherence of the weak coherent light of its use has seriously reduced picture quality, fuzzy useful edge in the image.Add the pathology in some patients' retina, in the OCT image, actual medical is used serious fuzzy, the phenomenon of rupture of important edges ubiquity inside and outside significant retina.The popularization of the three-dimensional OCT equipment of frequency domain in recent years, make the main flow of OCT retinal images provide for the form with three-dimensional data, but also there is no the border of can accurate, efficient and stable method and system extracting important retinal tissue layer in prior art.

Summary of the invention

The technical matters that (one) will solve

The technical problem to be solved in the present invention is: the border of accurate, the efficient and stable important retinal tissue layer of extraction how.

(2) technical scheme

For solving the problems of the technologies described above, the invention provides the extracting method on the three-layer laminated border of a kind of retina OCT volume data, comprise the following steps:

S1: retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

S2: each pixel by described difference volume data and described retina OCT three-dimensional data is calculated at the coordinate figure of vertical direction, to obtain border index volume data;

S3: search in the index volume data of described border, to obtain the divergent boundary point on current border;

S4: smooth operation is carried out in the set to described divergent boundary point, to obtain current border;

S5: judge whether to have obtained all borders of described retina OCT three-dimensional data, if, execution in step S6, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to step S1;

S6: all borders of exporting described retina OCT three-dimensional data.Preferably, in step S2, described border index volume data is calculated by following formula,

I _i，j，k＝w ₁*D _i，j，k+w ₂*Y _i，j，k

Wherein, Y _{I, j, k}Be retina OCT three-dimensional data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data.

Preferably, further comprising the steps of before step S2:

S0: described former retina OCT three-dimensional data is carried out level and smooth computing, to obtain the smooth body data;

In step S2, described border index volume data is calculated by following formula,

I _i，j，k＝w ₁*D _i，j，k+w ₂*S _i，j，k

Wherein, S _{I, j, k}Be smooth body data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data.

Preferably, all borders of described retina OCT three-dimensional data comprise: RPE layer border, IS/OS layer border and ILM layer border.

Preferably, in step S1, described differential filtering computing specifically comprises the following steps:

S11: centered by certain pixel of described retina OCT three-dimensional data, set up with other neighbor of described retina OCT three-dimensional data the square that a M * M * M pixel forms, described M is not equal to 1 positive odd number;

S12: the convolution operator corresponding according to current border, make the brightness value sum of the pixel on the center of described square deduct the brightness value sum of the pixel under the center of described square, and with the mean value of the subtraction value brightness value as the center of described square;

S13: with the center of other pixel as described square, return to step S11, until obtain the brightness value of all pixels of described retina OCT three-dimensional data, the brightness value of described all pixels consists of the difference volume data.

Preferably, the central point of establishing described square is Point on square except the center is Y _{I, j, k}, vertically more downward, when the k value was larger, the convolution operator of described RPE layer correspondence was:

$f_{i,j,k}=\left\{\begin{array}{c}1,k<k_0\\ 0,k=k_0\\ -1,k>{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\end{array}\right.$

The convolution operator of described IS/OS layer and ILM layer correspondence is:

$f_{i,j,k}=\left\{\begin{array}{c}-1,k<k_0\\ 0,k=k_0\\ 1,k>{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\end{array}\right.$

Preferably, in step S4, described smooth operation comprises the following steps:

S51: the gap of calculating the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of vertical direction of each divergent boundary point;

S52: in gap corresponding at least one divergent boundary point during greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations, return to step S51, if gap corresponding to described divergent boundary point be all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data.

Preferably, in step S5, when upgrading described retina OCT three-dimensional data according to current border, the brightness value of the current border in described retina OCT three-dimensional data and the pixel under current border all is made as zero.

The invention also discloses the extraction system on the three-layer laminated border of a kind of retina OCT volume data, comprising:

The differential filtering module is used for retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

Border index computing module is used for calculating at the coordinate figure of vertical direction by each pixel of described difference volume data and described retina OCT three-dimensional data, to obtain border index volume data;

Search module, be used for searching in described border index volume data, to obtain the divergent boundary point on current border;

Level and smooth module is used for smooth operation is carried out in the set of described divergent boundary point, to obtain current border;

Judge module, be used for judging whether to have obtained all borders of described retina OCT three-dimensional data, if, carry out output module, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to the differential filtering module;

Output module is used for all borders of exporting described retina OCT three-dimensional data.

Preferably, described smooth operation module comprises:

The gap calculating sub module is used for the gap to the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of the vertical direction of each divergent boundary point;

Judgement replaces module, when being used for gap at least one divergent boundary point correspondence greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations, return to the gap calculating sub module, if gap corresponding to described divergent boundary point is all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data.

(3) beneficial effect

The present invention includes following beneficial effect:

1. full-automatic: the present invention in extracting retina OCT volume data ILM, RPE and do not need artificial participation during the IS/OS coboundary, computation process has realized full-automation;

2. accuracy: the present invention can accurately extract ILM, RPE and the IS/OS coboundary in retina OCT volume data.

3. stable: the present invention is through a large amount of tests, and its algorithm is stable.

4. high efficiency: algorithm speed is fast, can mark three required boundary surfaces on ordinary individual's computing machine in more than ten second.

5. application is strong: according to three boundary surface positions that extract, can carry out thickness measure and visual to retina, and interactive three-dimensional is visual, has very strong clinical value at aspects such as retinal disease diagnosis.

Description of drawings

Fig. 1 is the process flow diagram according to the extracting method on the retina OCT volume data three-layer laminated border of one embodiment of the present invention;

Fig. 2 is the two-dimentional Bscan design sketch that obtains according to method shown in Figure 1;

Fig. 3 calculates according to method shown in Figure 1 boundary surface position and the visual retina gross thickness distribution plan that obtains that obtains.

Embodiment

Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.Following examples are used for explanation the present invention, but are not used for limiting the scope of the invention.

Fig. 1 is that the method for present embodiment comprises the following steps according to the process flow diagram of the extracting method on the retina OCT volume data three-layer laminated border of one embodiment of the present invention:

S1: retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

S2: each pixel by described difference volume data and described retina OCT three-dimensional data is calculated at the coordinate figure of vertical direction, to obtain border index volume data;

S3: search in the index volume data of described border, with the divergent boundary point that obtains current border (when searching, can be by the mode of A-scan, namely to described border index volume data I _{I, j, k}Each row in seek the pixel of brightness value maximum, described each classify I as _{I, j, k}Middle i and j be identical straight line all);

S4: smooth operation is carried out in the set to described divergent boundary point, to obtain current border;

S5: all borders that judge whether to have obtained described retina OCT three-dimensional data, if, execution in step S6, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to step S1 (in present embodiment, calculating successively RPE layer border, IS/OS layer border and ILM layer border);

S6: all borders of exporting described retina OCT three-dimensional data.

Preferably, in step S2, described border index volume data is calculated by following formula,

I _i，j，k＝w ₁*D _i，j，k+w ₂*Y _i，j，k

Wherein, Y _{I, j, k}Be retina OCT three-dimensional data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data;

In present embodiment, w ₁, w ₂All value is k.

Obtain precision for improving the border, preferably, further comprising the steps of before step S2:

S0: described former retina OCT three-dimensional data is carried out level and smooth computing, to obtain the smooth body data;

In step S2, described border index volume data is calculated by following formula,

I _i，j，k＝w ₁*D _i，j，k+w ₂*S _i，j，k

Wherein, S _{I, j, k}Be smooth body data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data;

In present embodiment, w ₁, w ₂All value is k.

Preferably, all borders of described retina OCT three-dimensional data comprise: RPE layer border, IS/OS layer border and ILM layer border, with reference to Fig. 2, be followed successively by from bottom to up RPE layer border, IS/OS layer border and ILM layer border in figure, with reference to Fig. 3, Fig. 3 is according to method shown in Figure 1 and calculates boundary surface position and the visual retina gross thickness distribution plan that obtains that obtains.

Preferably, in step S1, described differential filtering computing specifically comprises the following steps:

S11: centered by certain pixel of described retina OCT three-dimensional data, set up a M * M * M (in present embodiment with other neighbor of described retina OCT three-dimensional data, M=9) square of individual pixel composition, described M is not equal to 1 positive odd number;

S12: the convolution operator corresponding according to current border, make the brightness value sum of the pixel on the center of described square deduct the brightness value sum of the pixel under the center of described square, and with the mean value (described mean value is about to described subtraction value divided by M*M* (M-1)/2) of the subtraction value brightness value as the center of described square;

S13: with the center of other pixel as described square, return to step S11, until obtain the brightness value of all pixels of described retina OCT three-dimensional data, the brightness value of described all pixels consists of the difference volume data.

In present embodiment, the central point of establishing described square is

Point on square except the center is Y _{I, j, k}, vertically more downward, (also 1 and-1 value of convolution operator can be replaced mutually, subtract the value of top with the value below square) when the k value is larger, the convolution operator of described RPE layer correspondence is:

$f_{i,j,k}=\left\{\begin{array}{c}1,k<{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\\ 0,k=k_0\\ -1,k>{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\end{array}\right.$

The convolution operator of described IS/OS layer and ILM layer correspondence is:

$f_{i,j,k}=\left\{\begin{array}{c}-1,k<{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\\ 0,k={\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\\ 1,k>{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HDA0000086072900000022.png"\; state="\{\{state\}\}">k</figure-callout>}}_0\end{array}\right.$

Preferably, in step S4, described smooth operation comprises the following steps:

S51: the gap of calculating the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of vertical direction of each divergent boundary point;

S52: in gap corresponding at least one divergent boundary point during greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations (span of described default iterations is 1～10), return to step S51, if gap corresponding to described divergent boundary point be all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data

Preferably, in step S5, when upgrading described retina OCT three-dimensional data according to current border, the brightness value of the current border in described retina OCT three-dimensional data and the pixel under current border all is made as zero.

The invention also discloses the extraction system on the three-layer laminated border of a kind of retina OCT volume data, comprising:

The differential filtering module is used for retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

Border index computing module is used for calculating at the coordinate figure of vertical direction by each pixel of described difference volume data and described retina OCT three-dimensional data, to obtain border index volume data;

Search module, be used for searching in described border index volume data, to obtain the divergent boundary point on current border;

Level and smooth module is used for smooth operation is carried out in the set of described divergent boundary point, to obtain current border;

Judge module, be used for judging whether to have obtained all borders of described retina OCT three-dimensional data, if, carry out output module, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to the differential filtering module;

Output module is used for all borders of exporting described retina OCT three-dimensional data.

Preferably, described smooth operation module comprises:

The gap calculating sub module is used for the gap to the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of the vertical direction of each divergent boundary point;

Judgement replaces module, when being used for gap at least one divergent boundary point correspondence greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations, return to the gap calculating sub module, if gap corresponding to described divergent boundary point is all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data.

Above embodiment only is used for explanation the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (9)

1. the extracting method on a retina OCT volume data three-layer laminated border, is characterized in that, comprises the following steps:

S1: retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

S2: each pixel by described difference volume data and described retina OCT three-dimensional data is calculated at the coordinate figure of vertical direction, to obtain border index volume data;

Wherein: described border index volume data is calculated by following formula,

I _i，j，k=w ₁*D _i，j,k+w ₂*Y _i，j,k

Wherein, Y _{I, j, k}Be retina OCT three-dimensional data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data;

S3: search in the index volume data of described border, to obtain the divergent boundary point on current border;

S4: smooth operation is carried out in the set to described divergent boundary point, to obtain current border;

S5: judge whether to have obtained all borders of described retina OCT three-dimensional data, if, execution in step S6, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to step S1;

S6: all borders of exporting described retina OCT three-dimensional data.

2. the method for claim 1, is characterized in that, and is further comprising the steps of before step S2:

S0: described former retina OCT three-dimensional data is carried out level and smooth computing, to obtain the smooth body data;

In step S2, described border index volume data is calculated by following formula,

I _i，j，k=w ₁*D _i，j，k+w ₂*S _i，j,k

Wherein, S _{I, j, k}Be smooth body data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data.

3. method as described in any one in claim 1 ~ 2, is characterized in that, all borders of described retina OCT three-dimensional data comprise: RPE layer border, IS/OS layer border and ILM layer border.

4. method as claimed in claim 3, is characterized in that, in step S1, described differential filtering computing specifically comprises the following steps:

S11: centered by certain pixel of described retina OCT three-dimensional data, set up with other neighbor of described retina OCT three-dimensional data the square that a M * M * M pixel forms, described M is not equal to 1 positive odd number;

S12: the convolution operator corresponding according to current border, make the brightness value sum of the pixel on the center of described square deduct the brightness value sum of the pixel under the center of described square, and the mean value that obtains divided by M * M * (M-1)/2 with subtraction value is as the brightness value at the center of described square;

S13: with the center of other pixel as described square, return to step S11, until obtain the brightness value of all pixels of described retina OCT three-dimensional data, the brightness value of described all pixels consists of the difference volume data.

5. method as claimed in claim 4, is characterized in that, the central point of establishing described square is

Point on square except the center is Y _{I, j, k}, vertically more downward, when the k value was larger, the convolution operator of described RPE layer correspondence was:

$f_{i,j,k}=\left\{\begin{array}{c}1,k<k_0\\ 0,k=k_0\\ -1,k>k_0\end{array}\right.$

The convolution operator of described IS/OS layer and ILM layer correspondence is:

$f_{i,j,k}=\left\{\begin{array}{c}-1,k<k_0\\ 0,k=k_0\\ 1,k>k_0\end{array}\right.$

6. method as described in any one in claim 1 ~ 2, is characterized in that, in step S4, described smooth operation comprises the following steps:

S51: the gap of calculating the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of vertical direction of each divergent boundary point;

S52: in gap corresponding at least one divergent boundary point during greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations, return to step S51, if gap corresponding to described divergent boundary point be all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data.

7. method as described in any one in claim 1 ~ 2, it is characterized in that, in step S5, when upgrading described retina OCT three-dimensional data according to current border, the brightness value of the current border in described retina OCT three-dimensional data and the pixel under current border all is made as zero.

8. the extraction system on a retina OCT volume data three-layer laminated border, is characterized in that, comprising:

The differential filtering module is used for retina OCT three-dimensional data is carried out the differential filtering computing of vertical direction, to obtain the difference volume data;

Border index computing module is used for calculating at the coordinate figure of vertical direction by each pixel of described difference volume data and described retina OCT three-dimensional data, to obtain border index volume data;

Wherein: described border index volume data is calculated by following formula,

I _i，j,k=w ₁*D _i，j,k+w ₂*Y _i，j,k

Wherein, Y _{I, j, k}Be retina OCT three-dimensional data, D _{I, j, k}Be the difference volume data, k is that each pixel of described retina OCT three-dimensional data is at the coordinate figure of vertical direction, w ₁, w ₂Be the nonnegative real number that is directly proportional to vertical coordinate k, I _{I, j, k}Be border index volume data;

Search module, be used for searching in described border index volume data, to obtain the divergent boundary point on current border;

Level and smooth module is used for smooth operation is carried out in the set of described divergent boundary point, to obtain current border;

Judge module, be used for judging whether to have obtained all borders of described retina OCT three-dimensional data, if, carry out output module, otherwise, upgrade described retina OCT three-dimensional data according to current border, and upgrade the convolution operator of described differential filtering computing, return to the differential filtering module;

Output module is used for all borders of exporting described retina OCT three-dimensional data.

9. system as claimed in claim 8, is characterized in that, described smooth operation module comprises:

The gap calculating sub module is for the gap of the coordinate mean value of the vertical direction of other frontier points in the coordinate figure neighborhood default with it of the vertical direction that calculates each divergent boundary point;

Judgement replaces module, when being used for gap at least one divergent boundary point correspondence greater than threshold value, the coordinate figure of its vertical direction coordinate mean value with described vertical direction is replaced, and judge whether to have reached default iterations, if do not reach default iterations, return to the gap calculating sub module, if gap corresponding to described divergent boundary point is all less than threshold value or reach default iterations, with the set of the described divergent boundary point border as described retina OCT volume data.

Images